FIG. 1 and FIG. 2 illustrate a moving object simulation system employing virtual machines. A moving object simulation is a program that makes an imitative representation on a computer of how moving objects such as automobiles and trains travel on roads and railroads. Such a moving object simulation is used for predicting congestions at a time of an accident or for identifying places that cause chronic congestions.
FIG. 1 illustrates a map 2000 and moving objects 3000 that move on the map. With reference to FIG. 1, the map 2000 is divided into districts of a certain size, respectively named district 1, district 2, district 3, and district 4.
FIG. 2 illustrates an example employing virtual machines as computer environment for carrying out a moving object simulation. As illustrated in FIG. 2, physical machines 4000 to 4003 are respectively allocated to the districts described with reference to FIG. 1. The moving objects 3000 are caused to appear in each district by a moving object simulation program 120 operating on virtual machines. These virtual machines are managed by Hypervisors. In such a moving object simulation employing virtual machines, when a moving object 3000 moves from one district to another, a virtual machine on which the corresponding moving object simulation program 120 operates is migrated to one of the physical machines 4000 to 4003 that represents the district to which the moving object 3000 moves.
When a load on a physical server (physical machine 4000 to 4003) for a particular district exceeds a permissible amount due to a concentration of moving objects to the district (traffic congestion or the like), the number of physical servers allocated to the district needs to be increased. On the other hand, when the load on the physical server does not exceed the permissible amount, no physical servers need to be additionally provided. Estimating in advance whether the load will exceed the permissible amount enables appropriate additional provision of servers, for additional provision of physical servers entails calculation cost for a process of additional provision and financial cost for a the procurement of physical servers.
PTL 1 discloses a technique for disposing virtual machines on execution servers on the basis of a load calendar for a management of load prediction of virtual machines.
PTL 2 discloses a technique for making a moving plan for virtual machines more appropriately according to a state of network bands.
PTL 3 discloses a technique for executing a control on a physical server in accordance with a result of prediction of changes in a load on an object to be managed based on a history of load changes in the past.